Research on transformer fault diagnosis models with feature extraction

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Traditional transformer fault diagnosis algorithms suffer from low accuracy.
• Optimizing feature extraction and classification models is crucial for enhanced diagnostic performance.

Purpose of the Study:

• To develop a novel and highly accurate transformer fault diagnosis method.
• To leverage the Artificial Hummingbird Algorithm (AHA) for optimizing Kernel Principal Component Analysis (KPCA) and Extreme Learning Machine (ELM).

Main Methods:

• Utilized gas concentration ratio features for transformer fault diagnosis.
• Applied AHA to optimize KPCA kernel parameters, creating an AHA-KPCA feature extraction model.
• Employed AHA to optimize ELM input weights and hidden layer biases for an AHA-ELM classification model.
• Inputted AHA-KPCA extracted features into the AHA-ELM model for fault classification.

Main Results:

• The proposed AHA-KPCA-ELM method achieved a diagnostic accuracy of 95.73%.
• This accuracy surpasses traditional intelligent diagnostic methods and existing advanced algorithms.
• The method effectively extracts key features and classifies transformer faults with high precision.

Conclusions:

• The AHA-KPCA-ELM approach demonstrates superior effectiveness for transformer fault diagnosis.
• Optimizing KPCA and ELM with AHA significantly enhances diagnostic accuracy.
• This novel method offers a promising solution for reliable electrical transformer monitoring.